Wireless Pers Commun (2012) 67:573–584DOI 10.1007/s11277-011-0397-1

Very Simple BICM-ID Using Repetition Codeand Extended Mapping with Doped Accumulator

Khoirul Anwar · Tad Matsumoto

Published online: 13 September 2011© The Author(s) 2011. This article is published with open access at Springerlink.com

Abstract This paper proposes very simple bit-interleaved coded modulation with iterativedetection (BICM-ID) as a spectrally efficient signal transmission scheme, where irregularrepetition code and extended mapping with rate-1 doped accumulator (ACC) are combinedto achieve a clear turbo-cliff. Doped ACC is used for close matching of the demapper anddecoder’s extrinsic information transfer curves. Although the proposed BICM-ID systemis very simple, it can achieve excellent performances and completely eliminate bit-error-rate floor. An exemplifying result for extended non-Gray mapped 4-quadrature amplitudemodulation symbol shows that with 0.916 bits/channel use, turbo-cliff happens at 1.27 dBaway from the Shannon limit.

Keywords BICM-ID · Extended mapping · Iterative decoding · Repetition code

1 Introduction

Bit-interleaved coded modulation with iterative detection/decoding (BICM-ID) [1,2] hasbeen recognized as being a bandwidth efficient signal transmission scheme, where in thetransmitter, the encoder and the mapper are separated by an interleaver. Iterative detection-and-decoding takes place at the receiver by exchanging the extrinsic log-likelihood ratio

This research is supported in part by the Japan Society for the Promotion of Science (JSPS) Grant underScientific Research (C) No. 22560367 and (B) No. 23360170.

K. Anwar (B) · T. MatsumotoSchool of Information Science, Japan Advanced Institute of Science and Technology (JAIST),1-1 Asahidai, Nomi, Ishikawa 923-1292, Japane-mail: anwar-k@jaist.ac.jp

T. MatsumotoCenter for Wireless Communication, University of Oulu, 90014 Oulu, Finlande-mail: tadashi.matsumoto@ee.oulu.fi; matumoto@jaist.ac.jp

123

574 K. Anwar, T. Matsumoto

(LLR), obtained as a result of maximum a posteriori probability (MAP) algorithm fordemapping/decoding via interleaver/deinterleaver, following the standard turbo principle.

In principle, since the BICM-ID is a serially concatenated system, the analysis of itsperformance can rely on the area property of extrinsic information transfer (EXIT) chart.If the area between the EXIT curves of the demapper and the decoder is made very smallwhile keeping the tunnel open until a point very close to the (1,1) mutual information (MI)point, system can achieve performance closer to the Shannon limit with a clear turbo-cliff.Therefore, the transmission link design based on BICM-ID falls into the issue of matchingbetween the demapper and the decoder EXIT curves.

Various techniques have been proposed for better matching between the demapper anddecoder in BICM-ID without requiring heavy detection/decoding complexity aiming atachieving near-Shannon limit performance. One of the techniques that can provide us withdesign flexibility is extended mapping (EM). EM allocates more than m-bits to one signalpoint in 2m-quadrature amplitude modulation (QAM) format [1,3]. With EM, the demapperEXIT curve exhibits sharper decay, where the left-most point of the demapper EXIT curvehas a lower value than the demapper with Gray mapping, but the right-most point becomeshigher. With the sharper decay of the demapper EXIT curve, relatively weak codes such as amemory-1 convolutional code (CC) better matches the EM demapper than strong codes suchas Turbo codes [3].

References [1] and [4] propose the use of even weaker code, irregular repetition codes,where the degree distribution for the IRC with different variable node degrees (dvi ) in onecoded block are determined empirically to minimize the gap. Quite recently in [5], the authorshave shown that determining the optimal degree distribution falls into a simple convex opti-mization problem, solvable by using a linear programming technique.

Despite the simplicity of the BICM-ID system proposed by [1,4,5], however, it has afundamental drawback that it still suffers from the error floor. This is because the demapperEXIT curve does not reach a point very close to the (1,1) MI point.

In this paper, we use doped accumulator (ACC) [6] on the top of the BICM-ID structurepresented in [1,4], such that the EXIT curves are more closely matched, but do not intersecteach other, until a point close enough to the (1,1) MI point such that the EXIT trajectory cansneak-through the convergence tunnel. Another advantage of using doped ACC is that wecan achieve a higher spectrum efficiency by eliminating the necessity of the use of singleparity check code (SPC), which is necessary in [1] to reduce the error floor.

The system structure proposed in this paper is shown in Fig. 1. It is found that with thedoping ratio (1:P) = (1:1) the structure has a similarity to irregular repeat accumulate (IRA)1

coded BICM-ID with the check node degree dc being one, where � is a random interleaver.The smaller the P value the lower the left-most part of the EXIT curve of the combineddemapper and ACC, for which it is unavoidable that intersection between demapper and thedecoder EXIT curves happens at low MI point. Therefore, the doping ratio (1:P) should becarefully chosen. dc = 1 is acceptable because information bits are directly connected tothe extended mapper, by which LLRs of certain bits at the output of the demapper are notforced to be zero. Furthermore, dc = 1 does not sacrifice the spectrum efficiency in term ofbits-per-channel use because the EM rule, specified by μ, allocates more than m bits to onesignal constellation point.

A goal of this paper is not to present a capacity-approaching BICM-ID technique as in [7],but to show that with the proposed BICM-ID having a very simple structure, clear turbo-cliff

1 However, it is not similar to IRA when P > 1.

123

Very Simple BICM-ID Using Repetition Code 575

(without error floor) can still happen at a signal-to-noise power ratio (SNR) relatively closeto the Shannon limit for for a given code rate.

This paper is organized as follows. Section 2 proposes the structure of EM BICM-ID withdoped ACC and also describes the labeling pattern with several mapping schemes. Section 3analyzes the convergence properties of the proposed scheme using EXIT chart. Section 4evaluate the bit-error-rate (BER) performances of the proposed structure. Finally, Sect. 5concludes the paper with some remarks.

2 Proposed Simple Structure

2.1 Transmitter

As shown in Fig. 1, a binary bit information sequence b to be transmitted is encoded by IRCto obtain the coded sequence x , where the repetition times take several different values inone block, represented as dvi . By letting ai denote the distribution of the node degree dvi , thelength of IRC coded block is K = ∑

i ai · dvi × N , which is equal to the random interleaverlength, yielding the code rate of R = N/K , where N is the information length.

Repetition coded sequence x is bit-interleaved, resulting in time-permutated sequencey. The sequence y is doped accumulated with doping ratio of (1:P). The doped ACC is asystematic recursive convolutional code (SRCC) with coding rate R = 1. To keep the dopedACC’s code rate equal to one, the uncoded bit cu(k) at every k = n Pth bit is replaced bythe doped ACC-coded bit cc(k), which is the output of accumulated P interleaved repetitioncoded bits. With n = 1, 2, . . . , K/P and k being the bit index, the sequence c is expressedas

c(k) ={

cc(k) if k = n Pcu(k) otherwise.

(1)

With EM, � bits are mapped to one of the 2m constellation for modulation. Since � > mwith EM, more than one labels having different bit patterns in the segment are mapped onto each constellation point. However, there are many possible combinations of the labelingpatterns, and hence determining the optimal labeling pattern plays a crucial role towardsachieving near-Shannon limit performance.

This paper considers the optimized labeling which is extended non-Gray mapping suchthat with full a priori information, the MI between the IRC-coded transmitted bit and thedemapper output LLR is maximized. Figure 2a, b show the labeling pattern with � = 2 fornormal Gray and non-Gray mapping, respectively. Extended mapping with � = {3, 4, 5} [1]are shown in Fig. 2c, d, and e, respectively, where each 4-QAM symbol constellation pointhas 2�−m labels.

Fig. 1 Structure of EM BICM-ID with doped ACC and IRC

123

576 K. Anwar, T. Matsumoto

00

10

01

11

00

11

01

10

000101

011110

010111

011100

0100111000101000

1111010110010011

1101101101110001

1010110000000110

1001110110010111010101110011011111100111

1001000000110001000101010101000100101100

0010000001000101101011001100000100011100

1011101111001101110100101000111111011011

(a)Gray ( )

(b)Non-Gray ( )

(c) EM ( )

(d)EM ( ) (e) EM ( )

Fig. 2 Symbol labeling patterns for EM 4-QAM: a Gray (� = 2), b Non-Gray (� = 2), c EM � = 3,d EM � = 4, e EM � = 5

The ambiguity of bit identification at the receiver due to the 2�−m labels, allocated intoone constellation point, can be avoided by the use of a priori information provided by thedecoder, at each iteration, in the form of LLR [8]. The convergence behavior of the iterativedetection/decoding techniques can be evaluated using EXIT chart [8]. The impact of labelingpattern appears in the shape of EXIT curve representing relationship of a priori and extrinsicof MI between the coded sequence c and its corresponding LLR, as shown in Subsect. 2.4.

The � coded bits in sequence c are grouped to form a label for EM

q = [c1, c2, . . . , cv, . . . , c�], (2)

which is mapped to QAM symbol of

s = μ(q) (3)

according to the mapping rule μ(·). The EM-mapped symbols s is then transmitted over thechannel.

2.2 Channel

Additive white Gaussian noise (AWGN) channel having gain h is assumed. The receivedsignal is then expressed as

r = hs + η, (4)

where⟨|s|2⟩ = 1, and η is an independent and identically distributed (i.i.d) complex zero-

mean Gaussian random variable with variance σ 2 = 10−SNR[dB]/10.

2.3 Receiver

Iterative demapping and decoding process is invoked, where the extrinsic information isexchanged iteratively between the demapper and decoder. Decoding of ACC, referred toas ACC−1, is performed by using Bahl–Cocke–Jelinek–Raviv (BCJR) algorithm after EMdemapping (EM−1) and before de-interleaving �−1. Decoding of irregular repetition code

123

Very Simple BICM-ID Using Repetition Code 577

Fig. 3 Tanner graph of BICM-ID using IRC and EM with doped ACC

(IRC−1) is then performed and the iteration continues until no more relevant gains in extrinsicLLR can be achieved.2 When the convergence point is reached, binary decision is made onthe information bits b, based on the a posteriori LLR at each dvi . Obviously, keeping theconvergence tunnel open until the (1,1) MI point is necessary to achieve arbitrarily low BER.

With the help of a priori information La,M provided by the IRC−1 decoder, the demappercalculates the extrinsic LLR Le,M of the bit q[v] in (2) from the received signal r by

Le,M (q[v]) = lnP(q[v] = 0|r)

P(q[v] = 1|r)

= ln

∑

s∈s0exp

{

−|r − hs|2σ 2

} ∏�

w=1,w �=vexp{−q[w]La,M (q[w])}

∑

s∈s1exp

{

−|r − hs|2σ 2

} ∏�

w=1,w �=vexp{−q[w]La,M (q[w])}

, (5)

where s0, s1 indicates the set of labels in (3) having bit q[v] being 0 and 1, respectively.La,M (q[v]) is the a priori LLR fed back from the decoder IRC−1 corresponding to the vthbit position in the label q allocated to the signal point s.

The output extrinsic LLR of the demapper is then updated by the BCJR algorithm. LLRof every k = n Pth coded bit, cc(n P), is substituted by the LLR of the uncoded bit cu(n P)

which is obtained from the ACC−1, as shown in a Tanner graph of the proposed system inFig. 3.

The extrinsic LLR of demapper output, Le,M , is de-interleaved to obtain La,D . For the dvi

bits connected to one variable node, the extrinsic LLR is updated by summing up the LLRfor the (dvi−1) bits, as

Le,D( j) =dvi∑

j=1,j �= j

La,D(j), (6)

to produce extrinsic LLR of the j th IRC-coded bit.However, because IRC−1 does not produce LLR corresponding to the ACC-coded output

bits cc(n P), for the sake of complexity reduction, the feedback cc(n P) from the decoder tothe demapper is set at zero, as shown by Fig. 3.

2 In this paper, we perform at maximum of 50 iterations.

123

578 K. Anwar, T. Matsumoto

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1Interleaver: 38,500

SNR = 1 dB4-QAM with

Gray ( )

Extended Mapping (EM)

Standard Mapping

Fig. 4 EXIT curves of EM demapper for several �

2.4 Demapper and Mapping Pattern

Figure 4 compares for SNR = 1 dB the EXIT functions of 4-QAM demapper for several map-ping schemes. The X -axis, Ia,M = I (c; La,M ), is the a priori MI between coded sequencec and a priori LLR La,M provided by the decoder, while Y -axis, Ie,M = I (c; Le,M ), is theMI between c and its corresponding extrinsic LLR Le,M to be forwarded to the decoder.A mapping with � = 5 means that each 4-QAM symbol has 5 bits resulting in 8 labelingpatterns allocated to each constellation point.

It can be observed from Fig. 4 that the standard Gray mapping is flat, while standard non-Gray mapping (with � = 2) has decay. From the EXIT curves shown in Fig. 4, it is found thatstandard/non-extended mapping best matches turbo codes [9] and/or convolutional codeswith large memory, while the extended mapping (� > 2) is better matched with the repetitioncodes such as the code proposed in [1]. The matching analysis between the mapper andencoder is further discussed using EXIT analysis provided by Sect. 3.

3 EXIT Analysis of the Proposed Technique

With the Gaussianity assumption of the LLRs distribution, let the variance of LLRs La,M andLe,M , a priori and extrinsic LLR of demapper (including ACC−1), respectively, be denotedby σLa,M and σLe,M . The combined EXIT function of the demapper and ACC−1, denoted asTM , is then expressed by

Ie,M = TM (Ia,M,, SNR), (7)

where

Ie,M = J (σLe,M ) and Ia,M = J (σLa,M ). (8)

123

Very Simple BICM-ID Using Repetition Code 579

J (·) and J−1(·) are the functions that convert variance of LLR σL to its corresponding MI,and its inverse, respectively, given by [8]

J (σL) =(

1 − 2−H1σL2H2

)H3, (9)

J−1(Ia) =(

− 1

H1log2

(

1 − I1

H3a

)) 12H2

, (10)

where H1 = 0.3073, H2 = 0.8935, H3 = 1.1064, and σL = {σLa ,M , σLe,M }.The EXIT function of IRC−1, denoted as TD , is then given by

Ie,D = TD(Ia,D),

=∑

iai · dvi · J

(√(dvi − 1) · J−1(I i

a,D)2

)

∑

iai · dvi

, (11)

where I ia,D is the a priori MI of variable node with the i th degree and Ie,D is its output

extrinsic MI.Figure 5 shows for SNR = R · Eb/N0 = 1 dB the EXIT curves of the extended non-Gray

4-QAM demapper and the decoder used in [1,4,5], where 4-QAM mapping with � = 5 isassumed. The node degree distribution is dv1 = 5(a1 = 0.77), dv2 = 7(a2 = 0.23), withoutSPC.

According to the area property, the Shannon limit is approached by minimizing the gapbetween the two curves TM and T −1

D (defined in Eq. 11), while keeping the convergencetunnel open until a point very close enough to the (1,1) MI point. It is shown in [1] that withthe empirically obtained (near optimum) variable node degree distribution, the intersectionbetween TM and T −1

D is at Ia,M = Ie,D = 0.992. Since at high a priori MI the a posterioriLLR is almost the same as the extrinsic LLR, Ip,D ≈ Ie,D = 0.992. Hence, the bit-error-rate(BER) can be predicted as [8]

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.9 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

Decoder

Demapper with Acc

Demapper w/o Acc

SNR=1 dB, P=200R1=0.916 bits/channel use

dv1=5 (a1=0.77)dv2=7 (a2=0.23)

Fig. 5 EXIT charts of IRC without SPC and extended non-Gray 4-QAM with R1 = 0.916 bits/channel useat SNR = 1 dB and doping ratio of (1:200)

123

580 K. Anwar, T. Matsumoto

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.9 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 10.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

Decoder

Demapper with Acc

Demapper w/o Acc

SNR=3.1 dB, P=300R2=1.29 bits/channel use

dv1=3 (a1=0.58)dv2=5 (a2=0.41)dv3=6 (a2=0.01)

Fig. 6 EXIT charts of IRC without SPC and extended non-Gray 4-QAM with R2 = 1.299 bits/channel useat SNR = 3.1 dB and doping ratio of (1:300)

BER = 1

2erfc

(J−1(Ip,D)

2√

2

)

≈ 1.9 × 10−3 (12)

It is also suggested from (12) that the required MI to achieve BER of 10−5 is Ip,D = 0.99994.Ref. [1] proposes the use of SPC concatenated with IRC to push the T −1

D curve to the righthand side such that the intersection happens at a point close to (1,1) MI point. However, theuse of doped ACC with doping ratio of (1:200) bends the shape of the TM curve such that itcan reach a point very close to the (1,1) MI point, resulting in better matching between thetwo EXIT curves. Furthermore, the intersection of the two curves can be avoided, as shownin Fig. 5. Hence, the error floor can be completely removed. In this figure, we used the dopingratio of (1:200). It can be observed that the combined use of IRC and doped ACC makes thedemapper and decoder curves better matched than IRC alone for the BICM-ID technique.

The rate of IRC can be very flexibly changed by determining the distribution of dvi to pro-vide a turbo-cliff at a desired SNR. EXIT curves of an EM 4-QAM demapper and decoder,designed for turbo-cliff to happen at around SNR = 3.1 dB is shown in Fig. 6, of whichthe degree distribution is shown in the figure. Without doped ACC, the EXIT curves withthe demapper and the decoder intersect at around Ie,D = 0.995, where the predicted BER,according to (12), is about 1.136 × 10−3. However, the use of doped ACC with doping ratio(1:300) changes the shape of demapper EXIT curve to reach a point very close to (1,1) MIpoint, and hence no error floor is expected.

4 Performances Evaluation

A series of computer simulation was conducted to evaluate the performance and the conver-gence property of the proposed technique. To obtain reliable enough results, we transmitted100 blocks, each having 10,000 information bits. We consider random interleaver/deinterle-aver with a length of K = ∑

i ai · dvi × 10,000. The doping parameters for ACC is set

123

Very Simple BICM-ID Using Repetition Code 581

Table 1 Node degreedistribution for Case 1and Case 2

Case 1 dv1 = 5 (a1 = 0.77)

dv2 = 7 (a2 = 0.23)

Interleaver length K = 54,600

Case 2 dv1 = 3 (a1 = 0.58)

dv2 = 5 (a2 = 0.41)

dv3 = 6 (a3 = 0.01)

Interleaver length K = 38,500

-1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 410

-5

10-4

10-3

10-2

10-1

100

SNR (dB)

BE

R Proposed (w/o SPC)

Design in [1]

Shannon Lim

it = -0.52 dB

Information length: 10,000Interleaver: 54,600

R1= 0.916 bits/channel use

1.27 dB

P:400

P:200

Fig. 7 BER performances for R1 = 0.916 bits/channel use of extended non-Gray 4-QAM

to P = {200, 300, 400, 500}. In this paper we consider the degree distribution empiricallyobtained and used in [1,4] for extended 4-QAM with � = 5.

The performances are evaluated for two cases with the parameters chosen as in Table 1,which are the same as those used in the EXIT chart analysis presented in Sect. 3.

By eliminating the SPC codes and adding the rate-1 ACC with doping ratio of (1:P),the proposed EM BICM-ID provides better BER performance and higher bits/channel use,which are:

Case 1: R1 = �∑2

i=1 ai · dvi

= 0.916 (R1,Ref.[1] = 0.904) bits/channel use,

Case 2: R2 = �∑3

i=1 ai · dvi

= 1.299 (R2,Ref.[1] = 1.286) bits/channel use,

both with � = 5.Figure 7 shows, for comparison, BER performances with Case 1 of the proposed and

Ref. [1]’s techniques with maximum 50 iterations. The BER curve of EM BICM-ID withSPC exhibits minor improvement, however, the error floor still remains. On the contrary,the proposed technique with ACC and doping ratio of (1:200) or (1:400), can completelyeliminate the error floor even without the use of SPC. Assuming the Gaussian codebook, theShannon limit for R1 = 0.916 bits/channel use is

SNRlim(R1) = 10 log10(2R1 − 1),

= −0.52 dB. (13)

123

582 K. Anwar, T. Matsumoto

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 610

-5

10-4

10-3

10-2

10-1

100

SNR (dB)

BE

R Shannon Lim

it = 1.65 dB

Information length: 10,000Interleaver: 38,500

R2= 1.299 bits/channel use

Design in [1]

Proposed (w/o SPC)

1.61 dB P:300

Fig. 8 BER performances for R2 = 1.299 bits/channel use of extended non-Gray 4-QAM

BER performance for the Case 2 is presented in Fig. 8 for code rate R2 and doping ratio of(1:300) and (1:500), where the Shannon limit for R2 = 1.299 bits/channel use is

SNRlim(R2) = 10 log10(2R2 − 1),

= 1.65 dB. (14)

It can be observed from Figs. 7 and 8, that the proposed techniques improve not only thespectrum efficiency by removing the necessity of SPC but also the BER performance byeliminating the error floor while keeping the near-Shannon limit performance.

The performance improvements achieved by P = 400 and P = 500 for R1 and R2,respectively, with their degree allocations shown in Table 1 are the largest among othersettings tested. It should be noted here that instead of using empirically obtained degreedistribution, optimal node degree distribution dvi can be determined by the linear program-ming (LP) technique, such as that presented in [5], to further improve the performance of theproposed EM BICM-ID system. This is left as further research.

5 Conclusion

This paper has proposed a very simple BICM-ID technique using a combined use of IRCand doped ACC assisted extended non-Gray mapping. The proposed techniques provide: (a)higher spectrum efficiency by removing the necessity of SPC, (b) better matching betweenthe demapper and the decoder and hence eliminating the error floor caused by EXIT curvesintersection before they reach the (1,1) MI point. The numerical result for an exemplifyingdesign confirms that the turbo-cliff is achievable within 1.27 dB away from the Shannon limitfor 0.916 bits/channel use with 3-bit (� = 5) extended 4-QAM.

Acknowledgment We would like to acknowledge Dan Zhao for developing the simulation software of thedemapper part.

123

Very Simple BICM-ID Using Repetition Code 583

Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommer-cial License which permits any noncommercial use, distribution, and reproduction in any medium, providedthe original author(s) and source are credited.

References

1. Zhao, D., Dauch, A., & Matsumoto, T. (2009). BICM-ID using extended mapping and repetition codewith irregular node degree allocation. In IEEE VTC-Spring (pp. 1–5). Barcelona.

2. Li, X., & Ritcey, J. (1997). Bit-interleaved coded modulation with iterative decoding. IEEE Commu-nications Letter, 1(6), 169–171.

3. Henkel, P. (2006). Extended mapping for bit-interleaved coded modulation. In IEEE personal, indoorand mobile radio communications (PIMRC) (pp. 1–4). Helsinki.

4. Zhao, D., Dauch, A., & Matsumoto, T. (2009) Modulation doping for repetition coded BICM-ID withirregular degree allocation. In ITG wireless smart antenna. Berlin.

5. Fukawa, K., Zhao, D., Dauch, A., Tolli, A., & Matsumoto, T. (2010). Irregular repetition and singleparity check coded BICM-ID using extended mapping: Optimal node degree allocation. In ICSTconference on communications and networking in China (CHINACOM), Beijing. (Invited).

6. Pfletschinger, S., & Sanzi, F. (2006). Error floor removal for bit-interleaved coded modulation withiterative detection. IEEE Transactions on Wireless Communications, 5(11), 3174–3181.

7. Durisi, G., Dinoi, L., & Benedetto, S. (2006). eIRA codes for coded modulation systems. In IEEEICC (pp. 1125–1130). Istanbul.

8. ten Brink, S. (2001). Convergence behavior of iteratively decoded parallel concatenated codes. IEEETransactions on Communications, 49, 1727–1737.

9. Berrou, C., Glavieux, A., & Thitimajshima, P. (1993). Near Shannon limit error-correction coding anddecoding: Turbo code. In IEEE International conference on communication (ICC) (pp. 1064–1070).Geneva.

Author Biographies

Khoirul Anwar graduated (cum laude) from the department of Elec-trical Engineering (Telecommunications), Institut Teknologi Bandung(ITB), Bandung, Indonesia in 2000. He received Master and DoctorDegrees from Graduate School of Information Science, Nara Insti-tute of Science and Technology (NAIST) in 2005 and 2008, respec-tively. Since then, he has been appointed as an assistant professor inNAIST. He received best student paper award from the IEEE Radio andWireless Symposium 2006, California-USA, and Best Paper of Indo-nesian Student Association (ISA 2007), Kyoto, Japan in 2007. SinceSeptember 2008, he is with the School of Information Science, JapanAdvanced Institute of Science and Technology (JAIST) as an assistantprofessor. His research interests are network information theory, errorcontrol coding, iterative decoding and signal processing for wirelesscommunications.

123

584 K. Anwar, T. Matsumoto

Tad Matsumoto received his B.S., M.S., and Ph.D. degrees from KeioUniversity, Yokohama, Japan, in 1978, 1980, and 1991, respectively,all in electrical engineering. He joined Nippon Telegraph and Tele-phone Corporation (NTT) in April 1980. Since he engaged in NTT,he was involved in a lot of research and development projects, allfor mobile wireless communications systems. In July 1992, he trans-ferred to NTT DoCoMo, where he researched Code-Division Multiple-Access techniques for Mobile Communication Systems. In April 1994,he transferred to NTT America, where he served as a Senior Tech-nical Advisor of a joint project between NTT and NEXTEL Com-munications. In March 1996, he returned to NTT DoCoMo, wherehe served as a Head of the Radio Signal Processing Laboratory untilAugust of 2001; He worked on adaptive signal processing, multiple-input multiple-output turbo signal detection, interference cancellation,and space-time coding techniques for broadband mobile communica-tions. In March 2002, he moved to University of Oulu, Finland, wherehe served as a Professor at Centre for Wireless Communications. In

2006, he served as a Visiting Professor at Ilmenau University of Technology, Ilmenau, Germany, funded bythe German MERCATOR Visiting Professorship Program. Since April 2007, he has been serving as a Pro-fessor at Japan Advanced Institute of Science and Technology (JAIST), Japan, while also keeping the posi-tion at University of Oulu. Prof. Matsumoto has been appointed as a Finland Distinguished Professor for aperiod from January 2008 to December 2012, funded by the Finnish National Technology Agency (Tekes)and Finnish Academy, under which he preserves the rights to participate in and apply to European and Finn-ish national projects. Prof. Matsumoto is a recipient of IEEE VTS Outstanding Service Award (2001), NokiaFoundation Visiting Fellow Scholarship Award (2002), IEEE Japan Council Award for Distinguished Serviceto the Society (2006), IEEE Vehicular Technology Society James R. Evans Avant Garde Award (2006), andThuringen State Research Award for Advanced Applied Science (2006), 2007 Best Paper Award of Insti-tute of Electrical, Communication, and Information Engineers of Japan (2008), Telecom System Technol-ogy Award by the Telecommunications Advancement Foundation (2009), and IEEE Communication LettersExemplifying Reviewer Award (2011). He is a Fellow of IEEE and a Member of IEICE. He is serving as anIEEE Vehicular Technology Distinguished Lecturer during the term July 2011–June 2013.

123